ACCEL Researchers Earn Scientific Presentation Award at NISBRE Conference
Members of the DE CTR ACCEL Biostatistics, Epidemiology and Research Design (BERD) core attended and presented at the NISBRE 2024 Conference, held in Washington, DC, June 16 – June 19. Two of the presenters, Dr. Jobayer Hossain, director of biostatistics and senior research scientist at Nemours Children’s Health, and Hamed Fayyaz, graduate student in the computer sciences department at the University of Delaware, earned merit awards for their presentations. Details regarding both awards are included below.
Jobayer Hossain, PhD is a senior research scientist and director of the biostatistics program at Nemours Biomedical Research. He is a research professor of pediatrics at Thomas Jefferson University and an adjunct associate professor in the department of applied economics and statistics and adjunct faculty member in the center for bioinformatics & computational biology at the University of Delaware.
Presentation Title: Precision Race-Ethnic Disparity in Mortality Risk in Pediatric Cancers: A Study Using SEER Data
Abstract: Pediatric cancers exhibit significant heterogeneity in etiology, treatment response, and mortality. Mortality risk varies notably by histology-based cancer type, followed by race-ethnicity, posing a serious public health concern. Leveraging a comprehensive SEER dataset of 101,328 pediatric cancer patients diagnosed from 1975 to 2016, we employed frailty models to estimate mortality disparities among race-ethnic groups. Non-Hispanic African American (NHAA) and Hispanic patients faced significantly higher mortality risks than non-Hispanic Caucasians (NH-Caucasians), with NHAA experiencing the highest disparity. NHAA patients face higher mortality risks across all cancer types, with the greatest disparity (adjusted hazard ratio (aHR) between 1.50 and 2.00) in liver, endocrine, acute lymphocytic leukemia, urinary, chronic leukemia, and Hodgkin lymphoma. Moderate disparity (aHR between 1.20 and 1.49) is seen in the brain, CNS, eye, female genital system, AML, other leukemias, NHL lymphoma, and soft tissue including the heart. Low disparity (1.10-1.19) occurs in bone, joint, and adrenal gland cancers. In Hispanic patients, disparities vary slightly, with ALL and male genital system cancers showing the highest disparity, moderate disparity in 16 cancer types, and low disparity in AML and soft tissue and heart cancers. Hispanic pediatric patients exhibit a survival benefit in the digestive system compared to other groups. Racial/ethnic disparities persist across many cancer types, even after controlling for histological heterogeneity. NHAA patients experience the most severe survival disparity, followed by Hispanic patients. Understanding the estimated disparity in mortality risk can guide public health planners in addressing the modifiable gap in racial disparity in pediatric cancer mortality.
Synopsis: Pediatric cancers show significant heterogeneity in etiology, treatment response, and mortality. Mortality risk varies by histology-based cancer type and race-ethnicity. Non-Hispanic African American (NHAA) and Hispanic patients face higher mortality risks than non-Hispanic Caucasians (NH-Caucasians), with NHAA experiencing the highest disparity. Disparities persist across various cancer types, even after accounting for histological differences. Understanding this estimated disparity can guide efforts to address racial disparities in pediatric cancer mortality.
Funding: No specific funding is associated with this study. However, the abstract has been submitted as a DE-CTR investigator, given my role as a Senior Biostatistics Consultant at BERD.
—–
Hamed Fayyaz is a graduate student in the computer sciences department at the University of Delaware. The PI for this project is Dr. Beheshti from University of Delaware. Collaborators include Tim Bunnell from Nemours and Claudine Jurkovitz from ChristianaCare.
Title: An Interoperable Machine Learning Pipeline for Pediatric Obesity Risk Prediction Using EHR Data
Synopsis: Reliable prediction of pediatric obesity can offer a valuable resource to the providers, helping them engage in timely preventive interventions before the disease is established. Many efforts have been made to develop machine learning (ML)-based predictive models of obesity and some studies report high predictive performances. However, no largely used clinical decision support tool based on these ML models currently exists. This study presents a novel end-to-end pipeline specifically designed for obesity prediction, which supports the entire process of data extraction, inference, and communication via an API or a user interface. By using only routinely recorded data in electronic health records (EHRs), our pipeline uses a diverse expert-curated list of medical concepts to predict the risk of developing obesity. We have used input from various stakeholders, including ML scientists, providers, health IT personnel, health administration representatives, and patients throughout our design. By using the Fast Healthcare Interoperability Resources (FHIR) standard in our design procedure, we specifically target facilitating low-effort integration of our pipeline with different EHR systems.
Funding: Work supported by an Institutional Development Award (IDeA) from the National Institute of General Medical Sciences of the NIH under grant number U54-GM104941 (PI: Hicks) and the State of Delaware.